Postoperative pulmonary function was studied in 16 patients undergoing total hip or knee arthroplasty. Their mean age was 65 years. Half of them received spinal analgesia (22.5 mg bupivacaine + 0.3 mg morphine) and the other half underwent general anaesthesia with halothane-nitrous oxide. Four hours postoperatively, the forced expirogram was maintained in the spinal analgesia group, compared with preanaesthesia measurements. Functional residual capacity (FRC) measured by multiple breath nitrogen washout was reduced by 0.51, as was closing capacity (CC) measured by the bolus technique. The gas distribution index (nitrogen washout delay) was unaltered. The alveolar-arterial oxygen tension difference (PA-ao2) was not significantly altered, but arterial oxygen tension (Pao2) was reduced by 1.7 kPa, and arterial carbon dioxide tension (PaCo2) was increased. No respiratory measurements could be made in the general anaesthesia group 4 h postoperatively, but arterial blood gases were unaltered compared with preanaesthesia values. Eighteen hours postoperatively, forced vital capacity (FVC) was reduced in the spinal analgesia group, FRC and CC remained diminished and the gas distribution index was increased, indicating less efficient gas mixing. Simultaneously, PA-ao2 was increased, and Pao2 remained reduced despite increased alveolar ventilation (lowered PaCo2). In the general anaesthesia group FVC, FRC and CC were also reduced, but the gas distribution index remained at the awake level and blood gases were unaltered. It is suggested that the slight hypoventilation in the spinal analgesia group early after surgery may have contributed to impaired gas distribution and ventilation-perfusion matching later postoperatively.